For example, there is a case where a person performs a search to acquire information about a sightseeing spot that the person wants to visit at the time of creating a travel plan. In a case where there are a plurality of sightseeing spots that the person wants to visit, for example, the person searches a search site to acquire information about one of the sightseeing spots and then searches the search site again for information about another one of the sightseeing spots. In the search, a sightseeing spot name and another word are sometimes used in combination as search words.
As a method of acquiring a search word, for example, a method has been proposed of acquiring video data of a TV broadcast, extracting character information from regions in a predetermined number of frames between which there is no difference, and employing the extracted character information as a search keyword. For example, a method has been proposed of retrieving an image similar to an input recognition target image, extracting a plurality of keywords from information attached to the similar image, analyzing the extracted keywords, and specifying one of the keywords as a keyword that is likely to be a recognition result of the recognition target image. For example, Japanese Laid-open Patent Publication No. 2009-159437 and international Publication Pamphlet No. WO 2008/152805 are disclosed as related arts.
However, for example, the scope of a search becomes too large depending on a search word, and it may be difficult for a user to acquire desired information from a search result. In order to solve this problem, a method can be considered of storing the feature term of each Web site in a database to be used for a search and associating a sightseeing spot name and the feature term with each other. As a result, in a search on a sightseeing spot name, a web site having a feature term associated with the sightseeing spot name is hit. However, since feature terms include many conceptual words such as “explanation” and “experience”, an undesired search result may be listed high It is therefore desired that feature terms be classified.